Article Text
Abstract
Background: Declines in the prevalence of HIV might occur due to natural epidemic dynamics rather than changes in risk behaviour. Determining the cause of an observed decline is important in understanding the epidemiology of HIV.
Objective: To explore how patterns of recruitment and interactions between subpopulations in different areas influence the predicted decline in the prevalence of HIV in the absence of reductions in risk behaviour.
Methods: A deterministic mathematical model of the heterosexual transmission of HIV in high prevalence endemic settings incorporating various patterns of recruitment to high-risk behaviour groups, population growth and migration was solved numerically. The possibility that apparent trends are generated or obscured through aggregating data from across areas experiencing different epidemics is also considered.
Results: Declines in the prevalence of HIV can occur even if individuals do not change behaviour, raising the possibility that epidemic downturns could be wrongly attributed to interventions. This effect is greatest when individuals do not enter higher risk groups to compensate for reductions in size caused by deaths from AIDS and when migration is non-random with respect to risk or infectious status and migration patterns change as the epidemic matures. In contrast, aggregating prevalence data from subregions with different epidemic profiles tends to mask declines in prevalence.
Conclusions: Interpreting surveillance data is important in understanding widespread responses to HIV epidemics. The results show that understanding patterns of adoption of risk behaviours and patterns of migration is important in interpreting declines in the prevalence of HIV.
Statistics from Altmetric.com
Footnotes
Funding: PTW thanks the UK Medical Research Council; TBH thanks the Wellcome Trust; PJW thanks the Wellcome Trust and UNAIDS; GPG thanks UNAIDS, the Wellcome Trust and the UK Medical Research Council.
Competing interests: None.
The study was conceived jointly by the authors. The model was coded and analysed by PGW with the assistance of the other authors. All authors contributed to drafting the manuscript.